A PDE-based level-set approach for detection and tracking of moving objects

摘要: This paper presents a framework for detecting and tracking moving objects in sequence of images. Using statistical approach, where the inter-frame difference is modeled by mixture two Laplacian or Gaussian distributions, an energy minimization based we reformulate motion detection problem as front propagation problem. The Euler-Lagrange equation designed functional first derived flow minimizing then obtained. Following work Caselles et al. (1995) Malladi (1995), contours to be detected tracked are geodesic active evolving toward minimum energy, under influence internal external image dependent forces. level set formulation scheme Osher Sethian (1988), complex curves can topological changes naturally managed. To reduce computational cost required direct implementation, new approach exploiting aspects from classical narrow band fast marching methods proposed favorably compared them. In order further CPU time, multi-scale has also been considered. Very promising experimental results provided using real video sequences.